0

I have responses from individuals in the form of a Likert-type scale (range 1-7 where 1=strongly disagree and 7=strongly agree). There are 15 statements in 5 domains (3 statements per domain) relating to vaccine hesitancy, where importantly some statements are negatively worded and some positively worded. Of about 800 responders, about 10 gave the same response to all 15 statements (either strongly disagreed or agreed) and about 70 were always indifferent (all fours). In other words, their responses have no variation at all. Another group of about 20 gave conflicting responses within the same domain. That is, it seems they didn't read the statement carefully, since responses to each statement within the same domain should be correlated.

I'm wondering whether to exclude at least the first group from the analysis because it seems that their responses are irrational, but I don't know whether that's reasonable and if I do, what reference can I cite as justification. I've tried searching for keywords such as transitivity and monotonicity, and while these seem similar, they don't quite fit here.

Any help, including keyword(s) to search for or whether I should just include them all in the analysis, would be highly appreciated. :)

Edward
  • 206
  • I don't have any experience giving surveys, so take this with a grain of salt. I don't think you can throw any of them away because they are valid samples. If you suspect there is something wrong with your questions or sampling technique, then you should re-evaluate that to get more accurate responses. – Joff Jan 17 '22 at 04:56
  • Thanks @Joff. it was an online survey, to save costs and increase sample size, so cannot modify that part. The questions (statements actually) cannot be modified as they come from a validated scale. But I may do two separate analyses, one which includes all respondents and one which excludes some or all of the abovementioned groups as a sensitivity analysis and hope that results don't differ much. :) – Edward Jan 17 '22 at 05:26
  • One of the first things to learn about designing surveys is that some people act strarangely and others are unwilling or unable to make the effort to understand the questions. If you ask a question and then later its negatively worded equivalent, some people will think the second is just a repeat of the first and answer the same way as before. You can't expect participants to care as much about the topic of the survey as you do. – BruceET Jan 17 '22 at 09:29
  • google: survey data cleaning – frank Feb 09 '22 at 08:19

0 Answers0